One email a day: the top topics, each with the angle and the story, before you plan content. The daily report isn't open yet — leave your email and we'll tell you when it opens.
Image: unite.ai
Cooling
OpenAI says its models improved AMD's flagship GPU design by up to 14%
Heat 50 · #22 in AI · Cooling4 reports · 4 sources · First reported Nov 18 · Updated Nov 20
The story
OpenAI says its models rewrote SystemVerilog RTL for AMD's flagship GPU and improved power, performance and area by up to 14% against heavily optimized baselines, in under five months — roughly a year of progress, by OpenAI's account. Some model-written changes are already in a design AMD has sent to manufacturing, and OpenAI said AMD verified single-digit percentage gains across key blocks of its upcoming next-generation flagship CPU. The same approach on Jalapeno, OpenAI's first custom inference chip, cut standard-cell area by about 13% in the vector unit and 10% in the matrix unit, and the company is partnering with Synopsys to train a model for EDA tools. The figures are OpenAI's own: it gave no breakdown of the 14% across power, speed and area, and AMD has not commented.
Latest development
cooling per rule: score 70.0->50.0 (formula 70.0 - 20)
How it unfolded
cooling per rule: score 70.0->50.0 (formula 70.0 - 20)
the post carries no statement from AMD and 'up to 14%' is a maximum not broken down by power, speed or area. HN #9, 61 pts/18 cmt (post 50549493, links openai.com). Score 72.0 (rank 85, discussion 60, cross 70).
AMD also verified single-digit percentage PPA gains across key blocks of its upcoming next-generation flagship CPU from model-written designs. Same approach on Jalapeno, OpenAI's first custom inference chip: model-proposed changes cut standard-cell area about 13% in the vector unit and 10% in the matrix unit, proven in silicon (tapeout in nine months). On two open-source RISC-V cores, GPT-6.2 Astra scored highest across model generations (9,881 runs). OpenAI is also partnering with Synopsys to train GPT-Synopsys for EDA tools. Unite.AI and The Decoder (via mlq.ai) independently report the Nov 18 post. Caveat: figures are OpenAI self-reported
some changes are already in a design AMD sent to manufacturing, others added to production design files for future versions
Nov 18
OpenAI blog, 2026-11-18: its models rewrote SystemVerilog RTL on AMD flagship GPU IP and achieved up to a 14% improvement in power, performance and area (PPA) vs heavily optimized baselines in under five months, which OpenAI calls 'roughly a year of progress achieved in less than 5 months'
The angle
OpenAI's models just redesigned AMD's flagship GPU - and AMD shipped some of the changesSuggested format: Short video · for AI watchers, chip engineers, tech investors